For fifteen years the deal was simple: rank on page one, get the click. You optimized for a list of links, and the link was the prize.
That deal is quietly being renegotiated. More and more queries now end inside an AI-generated answer — Google's AI Overviews, ChatGPT with search, Perplexity — where the user gets what they need without clicking anything. The industry has a clunky name for adapting to this: Generative Engine Optimization (GEO). The name is new. The panic is real. Most of the advice about it is garbage.
Let me try to separate the signal.
The zero-click problem, stated honestly
Here's the uncomfortable shift, in one diagram:
Old funnel: query → ranked links → click → your page → conversion
New funnel: query → AI synthesizes an answer → (maybe) a citation → maybe a click
If the model answers the question inline, the click never happens. Your beautifully ranked page contributed to the answer and got nothing — no visit, no pixel, no attribution. This isn't hypothetical; it's already eating informational-query traffic.
The naive reaction is "SEO is dead, stop investing." That's wrong, and it's wrong in an expensive way. SEO didn't die — the unit of success changed from the click to the citation. You're no longer optimizing to be the link a human picks. You're optimizing to be the source a model picks when it composes the answer.
What actually gets a model to cite you
This is where I have to be careful, because nobody has the full ranking function and anyone who claims they do is selling something. But there are mechanics you can reason about from how retrieval-augmented systems work, and they're worth stating plainly.
1. Be chunk-retrievable, not just page-rankable. RAG systems don't ingest your page as a vibe. They split it into chunks, embed them, and retrieve the chunks most relevant to a query. A 3,000-word page where the answer is implied across five sections retrieves worse than a tight section that answers one question in one place. Structure for extraction:
## How much does X cost?
X costs $20–$50/month depending on tier. The free plan covers Y.
[self-contained. embeddable. quotable.]
A model can lift that as a citation. It cannot easily lift "as we discussed earlier, and depending on several factors..."
2. State claims as standalone, attributable facts. Models prefer to cite sources that say something concrete and checkable. "Studies show engagement improves" is unquotable mush. "In a 24-variant test on one skincare account, the AI-generated creative cut CPA by 18%" is a citable fact with a number attached. Specificity is retrieval bait.
3. Earn the entity association. AI systems build a graph of who-is-authoritative-about-what. Being mentioned, linked, and quoted across many sources for a topic makes you part of that topic's entity cluster — which is exactly what the model reaches for. This is the part you can't shortcut with on-page tricks; it's the slow compounding work of actually being a known source.
The trap: optimizing for the model and forgetting the human
Here's where a lot of GEO advice goes off a cliff. People hear "structure for machines" and start writing soulless, over-templated content stuffed with Q&A schema and FAQ blocks, hoping to game the answer box.
It backfires twice. First, the models are trained on human preference signals — content engineered to look machine-friendly often reads as low-quality and gets demoted. Second, even when you do get cited, a fraction of users still click through to verify or go deeper. If your page is robotic schema-bait, you've won the citation and lost the human. You're optimizing for two readers now — the model and the person — and they want more of the same thing than you'd expect: clarity, specificity, and a real point of view.
How to actually navigate this (without buying the hype)
The honest answer is that this is moving fast and the playbooks are mostly being written in real time by people doing it, not theorizing about it. My filter for what's worth reading: does the source show its work — real accounts, real costs, real failures — or does it just reword a press release?
The stuff that's held up for me has come from operators publishing actual experiments. I've found AI Marketing Focus useful precisely because it leans that way — head-to-head tool tests, costs and ROI shown openly, and a running thread on how AI search is reshaping SEO rather than vague "the future is AI" takes. Whatever sources you settle on, that's the bar: evidence over evangelism.
A concrete starting checklist
If you want to do something this week instead of doom-scrolling about it:
- Audit your top informational pages for chunkability. Can a single section answer a single question without context from the rest of the page? If not, restructure.
- Add specific, attributable facts — numbers, dates, named tools, real outcomes. Replace one piece of mush per page.
- Track citations, not just rankings. Query the AI engines for terms you care about and see who gets cited. That's your new SERP.
- Don't abandon classic SEO. AI engines still pull heavily from the same crawl-and-rank substrate. GEO is a layer on top, not a replacement.
- Write for a human who's skeptical. That instinct survives every algorithm change.
The actual takeaway
SEO didn't die. It got promoted into something harder: instead of persuading a ranking algorithm to show your link, you're persuading a language model that you're the most quotable, most credible source on a topic. The tactics will keep shifting. The underlying bet won't — be genuinely, specifically, verifiably useful, and structure that usefulness so a machine can find the exact piece it needs.
Everything else is a trench coat.
Are you tracking AI citations yet, or still living and dying by the ten blue links? And if you've found content engineered for the answer box that still reads well for humans — drop it in the comments, I want to see it.
Top comments (0)